JPH0612539B2 - Kanji / Kana conversion device - Google Patents

Description

TECHNICAL FIELD The present invention relates to a kanji-kana conversion system for automatically giving a reading kana to kanji in kanji-kana mixed data.

[Prior art]

In the conventional kanji-kana conversion method, the kana reading of kanji is basically used for each character, and the kanji reading is uniquely determined for each character. If the reading is other than that, an exception dictionary is prepared for the compound words that combine the characters before and after the kanji, and if the kanji to be converted in the conversion process is registered in the exception dictionary, that reading is prioritized. The method of letting is adopted. However, in this method, in a kanji character string in which several kanji compound words are combined, the reading is not always given in each compound word unit, so the part of the kanji character string that is not the originally intended compound word is not There is a possibility that a reading will be added. For example, when adding a reading to the idiom “development work”, the reading of each kanji will be open (kai), utterance (hatsu), work (sa), and kaku (guiyo).
If the idiom of seizure (hotssa) is registered in the exception dictionary, it will be mistaken for development work (kaihotsagiyo) and will be given a pseudonym.

On the other hand, when outputting mixed sentences of kanji and kana by voice, it is necessary to add reading kana of kanji, but when using it for such purposes, in order to add tones to the tones, kana It is desirable to make a mark. However, the conventional kanji / kana conversion method is based on the conversion on a character-by-character basis, and thus is not suitable for the above-mentioned applications.

In addition, as a method for assigning Kanji reading kana, the longest matching kanji (one kanji or two or more kanji compound words) registered in the kanji kana dictionary from the beginning of the kanji character string (word It is conceivable that the reading of the word base will be added and the reading of the word base will be added, but the wrong reading may be added due to the following factors.

(a) Occurrence of an affix or a idiom that has not been registered in the Kanji / Kana dictionary may cause incorrect recognition of the word base.

(b) The same compound word may have multiple readings.

[Object of the Invention]

An object of the present invention is to automatically give a reading kana to mixed kanji kana data, in a case where the kanji character string to which the reading kana is attached has a plurality of readings, and when there are a plurality of kanji character strings. An object of the present invention is to provide a method that can be divided and that can perform reading kana naming with extremely high accuracy when the divided character string is registered in a kanji kana dictionary.

[Outline of Invention]

According to the method of the present invention, recognition of a word base that is a unit of giving a kana,
The feature is that it is performed while confirming the selection of the reading kana in the following procedure.

(1) Kanji word base tentatively recognized by the longest match (forward word)
Based on the part-of-speech of the word base and the attributes of the preceding and following characters, the reading kana is given from the plurality of reading kana for the.

(2) For the Kanji character string following the word base tentatively recognized in (1), similarly, the word base (backward word) is tentatively recognized and its phonetic kana is given.

(3) Based on the number of characters of the front word and the rear word, the part-of-speech information, etc., based on the conversion processing control rule, the connection relation between the word bases is checked, and the validity of the word base division is checked to determine the reading.

For example, if a trailing word is identified as a single letter, if it is an affix (in this case, a suffix), then the leading word is considered to be split into the correct base, but if it is not, then the leading word is The word may not have been certified as a base having the originally intended meaning, and the base should be certified again.

Example of Invention

Hereinafter, the present invention will be described in detail with reference to the drawings. A hardware configuration of one embodiment of the present invention is shown in FIG. In FIG. 1, 1 is a processor, 2 is a storage memory for kanji / kana mixed data such as magnetic tape or magnetic disk, 3 is storage memory for kanji / kana converted kanji and its kana reading data, 4 is program storage memory, 5 is A work area, 6 is a kanji kana dictionary memory, 7 is a conversion processing control rule table memory, 8 is a reading kana giving rule table memory, and 9 is a line printer. The kanji / kana dictionary memory 6 stores a dictionary having a format as shown in FIG. That is, the kana character string A of one or more characters is used as an entry word, and its phonetic kana character string B, word attribute code C, and frequency information D are stored. If there are multiple readings for a given kanji character string, or if there are multiple word attribute codes for a given kanji character string and its reading kana character string, configure one set for each. Shall be taken. Word attribute code C
Is a code indicating the grammatical information of the Kanji character string A and its phonetic kana character string B, and the pronunciation of the pronunciation and the word attribute, and an example thereof is shown in FIG. The frequency information stores the kanji character string, its reading kana character string, and the frequency with which the word attribute code was used during kanji kana conversion.

The conversion processing control rule table memory 7 has a logical structure as shown in FIG. Each conversion process control rule is a combination of a condition and a process executed when the condition is satisfied.

Each row corresponds to each conversion processing control rule, and each column except the right end indicates the condition.

Yes, No, and-in the array indicate that each condition in each rule is (Yes) "must be satisfied", (N
o) Indicates that "it must not be satisfied", and (-) "any is acceptable". The rightmost column shows the contents of processing to be executed when the conditions of each rule are satisfied.

The smaller the rule number, the higher the priority of each rule.

The conversion processing control rule number 1 will be described as an example.

If the condition "Is the phonetic kana given to the forward word given?" Is satisfied and the condition "Is the character immediately after the forward word a kanji?" Is not satisfied, "Define the phonetic kana of the forward word and read it for the character string. The pseudonymization process is completed. "A physical example of the rule table (Fig. 4) is shown in Fig. 5. Here, "2" is Yes for each value in the array,
“1” 1 means No and “0” means Pass. Therefore, the values in rows a, b, and c in FIG.
It corresponds to a few conditions, respectively. Each value in FIG. 6 indicates a condition subroutine number for determining the condition corresponding to each column in FIG. For example, a subroutine number 1100 is added to the processing program for determining whether there is a Chinese character string of condition number 1. Each value in FIG. 7 indicates a processing sub-routine number executed when each condition is satisfied corresponding to each row in FIG. For example, "1" in FIG.
"0" indicates the subroutine number of the program that executes the process performed when all the conditions of rule number 1 in FIG. 4 are satisfied, that is, the process of "determining the front-back". The reading kana name giving rule table memory has a logical structure as shown in FIG.

The view and the physical implementation method of FIG. 8 are the same as those of the conversion processing control rule table.

The program has a module structure as shown in FIG.
In the Kanji / Kana conversion main module 20, a process of loading each rule table (FIGS. 4 and 81) into the processor 1, a process of extracting Kanji data to be subjected to Kanji / Kana conversion from the Kanji data storage memory 2, a Kanji data Is separated between non-Kanji and Kanji, and the process of setting the Kanji Kana conversion processing unit (called a phrase here), the process of referring to the conversion process control module for each processing unit, and the Kanji Kana converted Kanji The reading kana data is stored in the kana data storage memory 3
To output to the line printer and an error message for an error during execution of the process.

The conversion processing control module 21 refers to the conditional subroutine 22 corresponding to the subroutine number shown in FIG. 6 according to the conversion processing control rule table (FIG. 4), and the return code matches all the values in the rule table. The execution processing subroutine 23 of the subroutine number (see FIG. 7) corresponding to the rule having the smallest rule number is started. In the present embodiment, when the kanji character string to be converted into kanji kana is composed of a plurality of kanji character strings (called word bases) registered in the kanji kana dictionary, a mistaken recognition of the word base is prevented. For this reason, the method of assigning a phonetic alphabet for the word base while checking the validity by checking the connection relationship between two consecutive word bases (referred to as a forward word and a backward word). Therefore, the addition of the phonetic kana to the forward word and the addition of the phonetic kana to the backward word are also conditions for the conversion processing control rule. They are called a phonetic kana addition condition subroutine 22a for the forward word and a phonetic kana addition condition subroutine 22b for the backward word, respectively.

Similar to the conversion processing control module 21, each of the subroutines 22a and 22b is based on the reading kana provision rule table, and the condition subroutine 24 and the execution processing subroutine 2
5 and are started, and the optimum one is selected from a plurality of reading kana.

Next, the conversion processing procedure will be described by taking as an example the case of giving a reading kana to the kanji kana mixed data of "adding a new rule".

First, the processing contents of the kanji / kana conversion main module 20 will be described based on the flow chart of FIG.

When the Kanji / Kana conversion main module is started, the conversion processing control rule table (FIG. 4) is loaded from the conversion processing control rule table memory 7 onto the processor 1 and stored in the work area CRULE (101). The reading kana provision rule table (FIG. 8) is loaded from the reading kana provision rule table memory 8 onto the processor 1 and stored in the work area YRULE (102). Next, the data "addition of new rule" which is the target of kanji / kana conversion is read from the kanji / kana mixed data storage memory 2 (103). If the target data is not in the kanji / kana mixed data storage memory 2, the process of the kanji / kana conversion main routine is terminated. The read kanji / kana mixed data is divided at the change points from non-kanji to kanji and stored in the work area KANJI (104). In the above example, "new rule" is stored in the work area KANJI (1), and "additional" is stored in the work area KANJI (2).

The data obtained by dividing the kanji / kana mixed data is called a phrase here, and the number of the phrase is stored in the parameter NUM (105). In the above example, 2 is stored. 1 for K
The value 1 indicating the segment is stored (106), and the Kth segment KANJI (K) is stored in buffer A (107). In the above example, a kanji-kana conversion target character string is stored as shown in FIG. Store 0 in error code ERCODE (108)
After that, the conversion processing control module 21 is activated to perform the reading kana processing, and the result is stored in the buffer B (109).
In the above example, kanji and their readings are stored as shown in FIG. After processing, if the error code is not 0,
(110), the error message is output to the line printer (111), and the process of the Kanji-kana conversion main module is completed.

If the error code is 0 (110), the contents of buffer B are output to the kana data storage memory 3 (112). After adding 1 to K, it is judged whether there is a next clause (113) (1
14) If it exists, the processing from 107 onward is repeated, and if it does not exist, the processing from 103 onward is repeated. In the above example, with respect to the second phrase “addition”, the processing after 107 is performed, the reading kana is output to the kana data storage memory, and then 103, there is no kanji kana data to be converted to kanji kana. Therefore, the processing of the kanji-kana conversion main module is ended.

A conversion processing control method will be described based on the flow chart of FIG.

When the conversion processing control module 21 is activated, it stores 0 in the reprocessing counter AGAIN (201), stores 1 in I indicating the first conversion processing control rule (202), and stores the first in J. The conversion processing control condition of 1 is stored (20
3), 0 is stored in the return code (204), and the Jth conversion processing control condition subroutine is activated (20
5). It is determined whether an error has occurred during the processing of 205 (206), and if there is an error, the processing of the conversion processing control module 21 is ended. If there is no error, the value of the conversion processing control rule CRULE (I, J) is 0, or the return code RCOD of the processing result of the conversion processing control rules CRULE (I, J) and 205.
It is judged whether E and E are equal (207), and if satisfied, 208
If not satisfied, 210 is executed. 1 for J at 208
After adding (208), it is judged whether or not there is a Jth conversion processing control condition (209), and if satisfied, the processing after 204 is repeated. The execution processing subroutine is started (213). 21
Determine whether an error occurred during the processing of 3 (2
14) If there is an error, the processing of the conversion processing control module 21 is terminated, and if there is no error, it is judged in the processing of 213 whether a value other than 0 is stored in the reprocessing counter AGAIN (215). If the value of
The subsequent processing is repeated.

If the flag of reprocessing counter AGAIN is 0, buffer A
It is judged whether all the kanji stored in the kanji have been given readings (216). If the readings have been given to all kanji, the conversion process control module 21 ends the processing, and the kanji without readings are added. If there is, the processing after 201 is repeated.

At 210, after adding 1 to I, it is judged whether there is an I-th conversion processing control rule (211), and if satisfied, 20
Repeat the processing after 3 and if not satisfied, error code ER
1 is stored in CODE (212) and the process of the conversion process control module 21 is ended.

In the present embodiment, the conditional subroutine 2 of the conversion processing control rule
2, a subroutine (condition number 1 in FIG. 4) for determining whether the first character of the buffer A is a kanji character, a subroutine (condition number 2) for giving a reading to the forward word, and a character immediately after the forward word are kanji characters. Subroutine to judge (condition number 3)
And a subroutine for determining whether reprocessing is in progress (condition number 4)
And a subroutine (condition number 5) for determining whether the forward word has two or more characters, a subroutine (condition number 6) for determining whether the forward word is a prefix, and a subroutine for adding reading to the backward word (condition number 7). Subroutine for determining if the backward word is two or more characters (condition number 8) and subroutine for determining whether the backward word is a prefix or suffix (condition number 9)
Then, there is a subroutine (condition number 10) that determines whether the character immediately after the backward word is Kanji. If the condition is satisfied after executing the process, store 2 in the return code RCODE and satisfy the condition. For example, 1 is stored in the return code RCODE.

Further, as the execution processing subroutine 23 of the conversion processing control rule, the forward word in the buffer E and its reading kana are buffer B.
In the buffer E (see the processing of rule number 1 in FIG. 4), and a subroutine for storing the forward word and its reading kana in buffer E and the backward word in buffer F and its reading kana in buffer B (rule number 2). , 7 and 11), the forward word in the buffer E and its phonetic kana are stored in the buffer B, and the character string without the phonetic kana in the buffer A is stored in the buffer A.
Subroutine to move to the beginning of (rule number 3, 6, 8, 1
2), the forward word of the buffer E and its phonetic kana and the backward word of the buffer F and its phonetic kana are stored in the buffer B,
Subroutine (rule number 4, 9) for moving the character string without reading kana in buffer A to the beginning of buffer A and adding 1 to the reprocessing counter AGAIN, and the forward word of buffer E and its reading kana and the back of buffer F Second word and its phonetic kana
As shown in FIG. 0, a subroutine (rule numbers 5 and 10) for storing in buffer G, a kanji character string of buffer G and its phonetic kana are stored in buffer B, and a character string not yet given in phonetic A in buffer A To the beginning of buffer A (rule numbers 13 and 14), and if there is a non-Kanji character string in buffer A, there is a subroutine to store it in buffer B (rule number 15).

If an irrational event occurs during processing, a value other than 0 is stored in the error code ERCODE and the processing ends.

In the “new rule” of the above example, the idiom “new rule” is not registered in the kanji kana dictionary 6, but “new”, “rule”, “new”, “rule”, and “rule” are registered. Then, in the first time, the conversion processing control rule number 5 shown in FIG. 4 is applied, the reading kana is given by “new” and “rule”, and it is temporarily stored in the buffer G, Processing counter AGAIN
1 is stored in. Since the reprocessing counter is not 0 (215), the conversion process is performed again based on the conversion process control rule, the conversion process rule number 11 is applied, and "new" is applied.
And a "rule" are added to the reading kana and stored in the buffer B (see FIG. 15). After that, the process of the conversion process control module 21 ends.

For the above example "Add", add "Add" to the Kanji Kana dictionary.
In the first time, the conversion processing control rule number 1 is applied, the phonetic kana “Tsuika” is added to “Add”, the result is stored in the buffer B, and then the conversion processing control module 21 is added. The process ends.

Next, a method of giving reading kana to the forward word and the backward word will be described with reference to the flowcharts of FIGS. 12 and 13.

As shown in FIG. 12, the reading kana assigning subroutine for the forward word, when activated, stores 0 in the rematch counter FAGAIN (1101) and determines whether the reprocessing counter AGAIN is 0 (1102). , 0, the processes after 1104 are executed, and if it is not 0, the value of the re-matching counter AGAIN is assigned to the re-processing counter FAGAIN (1103), and then 110
The processing after 4 is executed. At 1104, the FI first
After storing 1 indicating the 1st reading kana provision rule (see FIG. 8) (1104) and storing 1 indicating the 1st reading kana provision condition in FJ (1105), 0 is set to the return code YRCODE. It is stored (1106) and the FJ-th reading kana provision condition subroutine is activated (110).
7).

It is determined whether an error has occurred during processing 1107 (11
08), if there is an error, the processing of the reading kana addition subroutine for the forward word is terminated, and if there is no error, the value of the reading kana addition rule YRULE (FI, FJ) is 0 or YRULE (FI, FJ)
And the return code YRCODE of the processing result of 1107 are equal (1109), and if satisfied, 111
The processing after 0 is executed, and if not satisfied, the processing after 1112 is executed.

At 1110, after adding 1 to FJ (1110), the Fth
Judge whether there is a Jth reading kana provision rule condition (1
111), if satisfied, repeat the processing after 1106,
If not satisfied, the FIth reading kana addition execution processing subroutine is started (1115). In 1115, the added reading kana data is as shown in FIG.
It is stored in the buffer D.

It is determined whether or not an error has occurred during the processing of 1115 (1116), and if there is an error, the processing of the reading kana assigning subroutine for the preceding word is ended. If there is no error, the data of buffer D is After moving to the buffer E (1117) as illustrated in the figure, in the processing of 1115, it is determined whether or not a value other than 0 is stored in the re-verification counter FAGAIN (1118). If it is stored, the processing from 1102 onward is repeated, and if 0 is stored, the reading kana addition processing for the forward word is ended.

At 1112, after adding 1 to the FI (1112), the FI
It is judged whether there is the second reading kana assignment rule (111
3) If it exists, the processing after 1105 is repeated. If it does not exist, 2 is stored in the error code ERCODE and the processing of the reading kana addition subroutine for the preceding word is completed.

When the reading kana addition subroutine for the backward word is activated as shown in FIG. 13, the rematch counter BAGAIN is set to 0.
Is stored (1201), the parameter BI is set to 1 indicating the first reading kana provision rule (1202), and the parameter BJ is set to 1 indicating the first reading kana provision condition (1203). , 0 is stored in the return code YRCODE (1204), and the BJ-th reading kana provision condition subroutine is activated (1205). During processing of 1205, it is judged whether an error has occurred (1206), and if there is an error, the processing of the reading kana addition subroutine for the backward word is ended. If there is no error, the reading kana addition rule YR
The value of ULE (BI, BJ) is 0, or YRULE (BI, BJ) is 120
It is determined whether the value of the return code YRCODE of the processing result of 5 is equal (1207), and if satisfied, the processing of 1208 and subsequent steps is executed, and if not satisfied, the processing of 1211 and subsequent steps is executed.

In 1208, after adding 1 to BJ (1208),
It is determined whether or not there is a J-th reading kana provision condition (120
9) If it is satisfied, the processing from 1204 onward is repeated, and if it is not satisfied, the BI-th reading kana addition execution processing subroutine is started (1210). At 1210, the added reading kana data is stored in the buffer D.

It is determined whether or not an error has occurred during the processing of 1210 (1214), and if there is an error, the processing of the reading kana assigning subroutine for the backward word is ended. If there is no error, the data of buffer D is set to the 19th. As shown in the figure, after moving to the buffer F (1215), it is judged in the process of 1210 whether a value other than 0 is stored in the rematch counter BAGAIN (1216). If it is stored, the processing from 1202 onward is repeated, and if 0 is stored, the phonetic provision processing for the backward word is ended.

At 1211, after adding 1 to BI (1211), the BI
It is determined whether there is the second rule for assigning reading kana (121
2) If it exists, the processing after 1203 is repeated, and if it does not exist, 3 is stored in the error code ERCODE and the processing of the reading kana giving subroutine for the backward word is ended.

In this embodiment, as a conditional subroutine of the rule for assigning phonetic kana, among the kanji character strings registered in the kanji kana dictionary, the one that has the longest match from the beginning of the kanji kana converted character string of buffer A or if If the collation flag is a value other than 0, the number of characters obtained by subtracting the value of the re-collation flag from the number of longest matching characters is taken out, stored in buffer C shown in FIG. 16, and it is determined whether or not it is a character string. The subroutine (see FIG. 8, condition number 1) and the subroutine (condition number 2) that determines whether the Kanji character string of buffer C is one character, and the Kanji character string of buffer C can be identified as a five-stage verb. There is a subroutine (condition number 3) that determines whether there is something, and a subroutine (condition number 4) that determines whether or not there is one that can be recognized as a prefix in the Kanji character string of buffer C. Subroutine (condition number 5) that determines whether or not there is a kanji character string that can be recognized as a suffix, and subroutine that determines whether or not any of the kanji character strings of buffer C can be recognized as an adjective ( Condition number 6) and the Kanji character string of Bathua C
Subroutine (condition number 7) that determines whether there is anything other than a prefix, suffix, adjective, on-reading, and kun-yomi, and a subroutine that determines whether the Kanji character string of buffer C has both on-yomi and kun-yomi There is a subroutine (condition number 9) that determines whether the character immediately after the longest matching character string in buffer A is Kanji (condition number 9), and if the condition is satisfied after the processing is executed, the return code YRCODE To 2
Is stored, and if the condition is not satisfied, 1 is stored in YRCODE.

Also, as an execution processing subroutine of the reading kana provision rule,
A subroutine (a process corresponding to the rule numbers 1 and 7 in FIG. 8) for storing a character whose word attribute is a five-stage verb in the buffer C's Kanji character string and the buffer C's Kanji character string Among these, a subroutine (rule number 2, 8) for storing the one having the word attribute as the prefix in the buffer D and a subroutine (rule number 3, for storing the one for which the word attribute is the suffix as the suffix) 9) and a subroutine (rule numbers 4 and 10) for storing in the buffer D the character whose word attribute is an adjective in the Chinese character string of buffer C and the word attribute in the Chinese character string of buffer C. Is a five-verb, prefix,
Subroutine for storing in the buffer D anything other than suffixes, adjectives, on-reading and kun-reading (rule numbers 5 and 11)
And a subroutine (rule number 6) for adding 1 to the re-matching flag, and a subroutine (rule number 12) for storing in the buffer D a Kanji character string of the buffer C whose word attribute is on reading. In the character string of Bathua C,
There is a sub-routine (rule number 13) for storing the word attribute of kun-yomi in buffer D, and a subroutine (rule number 14) for storing the word attribute of on-reading or kun-reading in buffer D. During the processing of each execution processing subroutine, if there is more than one to be stored in the buffer D, the one having the highest frequency of each Kanji character string in the buffer C is selected, and if the frequency is the same, the buffer C is selected. The one located in the upper row shall be selected.

Also, if an unreasonable event occurs during processing of each condition subroutine and each execution subroutine, an error code
Store a value other than 0 in ERCODE and end the process.

〔The invention's effect〕

 According to the present invention, the following effects can be obtained.

(1) You can add reading kana in word base units to the kanji character strings of mixed kanji kana sentences.

(2) By using the method of registering the phonetic kana selection criteria for a kanji character string having a plurality of kana characters as a rule with a set of conditions and execution processes, rules can be changed or added / deleted. It takes the form of table correction, addition / deletion of conditional subroutines, addition / deletion of execution subroutines, and since it is not necessary to change the conversion processing procedure, the rules can be easily improved and expanded.

(3) The conversion processing procedure can also be easily improved and expanded by formulating it as a conversion processing control rule.

(4) By separating the rule for controlling the conversion process and the rule for giving the reading kana, the number of kanji and kana conversion rules can be reduced.

In the embodiment, the conversion processing control rules are 15 and the reading rules are 14 in total, 29 rules in total, but if they are not separated, the reading words are given to the forward and backward words. From about 100 times as many as 14 × 14 × 15 = 2940 rules are required.

[Brief description of drawings]

FIG. 1 is a hardware configuration diagram of a kanji / kana conversion system according to an embodiment of the present invention, FIG. 2 is a diagram showing items of a kanji / kana dictionary, and FIG. 3 is a word which is one item of a kanji / kana dictionary. FIG. 4 is a diagram showing an example of attributes, FIG. 4 is a diagram showing an example of a logical configuration of a control rule for Kanji-Kana conversion processing, and FIG. 5 is a diagram showing how the Kanji-Kana conversion control rule is stored in the memory. 6 and 6 show how the subroutine numbers corresponding to the respective conditions of the Kanji-Kana conversion control rule are stored in the memory, and FIG. 7 corresponds to each execution process of the Kanji-Kana conversion control rule. FIG. 8 is a diagram showing how the numbers of the subroutines are stored in the memory, FIG. 8 is a diagram showing a logical configuration example of the reading kana naming rule, and FIG.
FIG. 10 is a diagram illustrating a software module configuration of an embodiment of the present invention, FIGS. 10 (A) and (B) are diagrams showing a processing procedure of a kanji / kana conversion main module, FIG. 11 (A), (B)
Is a diagram showing a processing procedure of the conversion processing control module based on the conversion processing control rule, and FIGS. 12 (A) and 12 (B) are diagrams showing a processing procedure of giving the pronunciation of a forward word based on the reading kana character assignment rule. , FIG. 13 (A) and FIG. 13 (B) are diagrams showing a processing procedure for assigning the pronunciation of a backward word based on the rules for assigning the pronunciation kana, FIG. 14 is a diagram illustrating the configuration of the buffer A, and FIG. FIG. 16 illustrates the configuration of buffer C, FIG. 17 illustrates the configuration of buffer C, FIG. 17 illustrates the configuration of buffer D, and FIG. 18 illustrates the configuration of buffer E. FIG. 19 and FIG. 19 are views exemplifying the structure of the buffer F, and FIG.
FIG. 0 is a diagram illustrating the configuration of the buffer G.

Front page continued (72) Inventor Junji Tsuda 1099 Ozenji, Aso-ku, Kawasaki-shi, Kanagawa Inside the Hitachi, Ltd. System Development Laboratory (56) References JP-A-56-92677 (JP, A)

Claims (4)

[Claims] 1. A dictionary in which a word consisting of a kanji character string and at least one or more reading kana character strings and word attributes of the word are stored in association with each other, and the word attribute of the word consisting of a kanji character string can be recognized. First storage means for storing a first rule consisting of a condition and a process of selecting a reading kana of a word consisting of the kanji character string in accordance with the condition of the satisfaction of the condition, and at least two input kanji character strings. Means for temporarily dividing into two words and giving a reading kana character string to the words temporarily divided based on the dictionary and the first rule; word attributes of two consecutive words in the temporarily divided kanji character string; Second storage means for storing a second rule consisting of a condition relating to whether or not to give a reading kana, a process of determining a reading kana character string to be executed corresponding to the condition of the satisfaction of the condition, and a process of redoing the temporary division. And the above Means for determining the satisfaction of the condition defined by the second rule for two words, checking the validity of the division of the two consecutive words with reference to the second rule, and dividing by the checking means. And a means for determining the reading kana character string based on the second rule when it is determined to be valid. 2. The kanji / kana conversion device according to claim 1, wherein the means for assigning the kana character string includes the means for assigning the kana when the checking means determines that the tentative division is inappropriate. Based on the processing prescribed in 2 rules,
A kanji / kana conversion device, wherein the temporary division is performed again. 3. The kanji-kana conversion device according to claim 1, wherein the first rule is associated with a plurality of conditions relating to approval or disapproval of word attributes of words, and a condition in which the plurality of conditions are satisfied. An apparatus for converting kana to kana, comprising a plurality of processes, wherein the first storage means stores the first rule in a table format. 4. The kanji / kana conversion device according to claim 1, wherein the second rule includes a plurality of conditions regarding word attributes of two consecutive words and whether or not to give a reading kana.
A kanji / kana conversion device, comprising a plurality of processes associated with the established condition of the plurality of conditions, wherein the second storage means stores the second rule in a table format.